[

CYCLIC BEHAVIOR OF BEAM-TO-COLUMN

WEAK-AXIS ·MOMENT CONNECTIONS

by

Kenneth A. Heaton

FRITZ ENGINEERING l:ABORATORY LIBRARY

A Thesis

Presented to the Graduate Committee

of Lehigh University

in Candidacy for the Degree of

Master of Science

m

Civil Engineering

Lehigh University

Bethlehem, Pennsylvania

May 1Q87

Acknowledgments

The analytical and experimental study presented m this thesis was

conducted at Fritz Engineering Laboratory, Lehigh University, Bethlehem,

Pennsylvania. Dr. Irwin J. Kugelman is chairman of the Department of Civil

Engineering.

The experiments conducted in this work form part of the investigations

"Cyclic Behavior of Moment Connections" sponsored by the National Science

Foundation (Grant No. ECE-8320540).

The author wishes to express his sincere thanks to Dr. George C. Driscoll,

supervisor of the thesis and related research work, for the benefit of his helpful

suggestions and encouragement, patience and guidance during the course of this

study, and demonstrating to the author both the objectivity of science and the

science of humor.

Many thanks are due to Dr. Le-Wu Lu, Dr. Lynn S. Beedle, and Dr.

Bruce R. Somers for their valuable ideas and suggestions during the different

phases of the research project. Many people helped in the preparation of

specimens, erection of the assemblage, and in the actual testing, especially Dr.

Roger G. Slutter, and Messrs. Hugh T. Sutherland, Robert R. Dales, Charles

F. Hittinger, Raymond Kromer, Russell Longenbach, Daniel Pense, Dave Kurtz,

Gene Matlock and Todd Anthony. Mr. Richard N. Sopko took all test

photographs and prepared the prints and slides. The assistance of Ms. Eleanor

Nothelfer in obtaining various articles will never be forgotten.

11

Table of Contents \.

Acknowledgments ii Abstract 1 1. Introduction 2 2. Considerations in the Design of the Specimen 5 3. Considerations in the Design ot the Test Set-Up 9

4. Results of the Test Program 13

5. Test Data Management 18

6. Computer Modeling 23

1. Discussion of Results 26

8. Conclusions 37

References 40

Appendix A. Tables & Figures 41

Vita 75

111

Figure A-1: Figure A-2: Figure A-3: Figure A-4: Figure A-5: Figure A-6: Figure A-7: Figure A-8: Figure A-9: Figure A-10: Figure A-11: Figure A-12: Figure A-13: Figure A-14: Figure A-15: Figure A-16: Figure A-17: Figure A-18: Figure A-19: Figure A-20: Figure A-21: Figure A-22: Figure A-23: Figure A-24: Figure A-25: Figure A-26: Figure A-27: Figure A-28: Figure A-29: Figure A-30: Figure A-31:

List of Figures

Typical Connection Detail Type 1 Connection Detail Type 2 Connection Detail Reaction Frame & Test Setup Instrumentations for Tests Loading Sequence for Test 1 Loading Sequence for Test 2 Loading Sequence for Test 3 Load-Deflection for Test 2

Load-Deflection for Test 3 STRUCTR Model for Connection Members Exploded Force Diagram for Connection Members Bolt Slip at Edges of Web Plate Fixed End Beam Supports for Top Flange Plate Simple Span Bending of Top Flange Plate Deflection at Column Flange Tips Shear Stress Between Rosettes 12 & 15 Shear Stress Between Rosettes 1 & 2 Shear Stress Between Rosettes 5 & 6 Shear Stress Between Rosettes 8 & 9 Bending Stress Between Rosettes 1 & 2 Bending Stress Between Rosettes 8 & 9 Stress Flow from Web to Flanges Rosette Location of Bottom Flange Bending Stress Between Rosettes 20 & 29 Integration of Rosettes for Web Shear Shear Stress Distribution in Beam Web Through-Thickness Restraint on Welds Rosette Location on Top Flange Rosette Location on Beam Web Proposed Suggestions for Improvement

lV

44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74

. '

Table A-1: Table A-2: Table A-3:

List of Tables

Comparision of STRUCTR and Test Results Comparision of Stresses for Rl & RS Comparison of STRUCTR with Measured Loads

v

42 43 43

Abstract

KEY WORDS: columna(support);cyclic loading(earthquakes),framed structures,finite element method(analytical model);plasticity; fracture mechanics;welded joints,matrix methods.

Continuation of the beam-to-column research begun at Fritz Engineering

Laboratory, Lehigh University simulating the cyclic loading of a beam-to-column

weak-axis moment connction. Behavior of this sort is required to determine the

acceptability of a fully welded design when subjected to a dynamic loading.

Critical factors m the connection detail are the thickness of the connection

plate, weld sizes, elimination of a backing stiffener opposite the beam and

defects in workmanship. Discussion of previous work on full scale moment

connections, considerations in the design of the test set-up and the overall

approach to the intent of the testing procedure are expressed. Micro-computer

based data reduction programs are implemented to retrieve and graphically

display the data. Theoretical studies are presented as . a means to relate

empirical results to matrix structural design methods to be considered in· future

building codes. Presentation of experimental results concerning ultimate load

and ductility for two types of welded details with conclusions regarding areas

for future study.

REFERENCE: Heaton, Kenneth A.,"Cyclic Behavior of Beam-to-Column Weak-Axis Moment Connections,• Master's Thesis, Lehigh Univ., May, 1987.

1

Chapter 1

Introduction

Over the past forty years a senes of research projects on the subject of

beam-to-column connections were conducted in the Fritz Engineering Laboratory

at Lehigh University. Previous work included riveted connections, semi-rigid

connections, flexible welded angle connections, subassemblages representing a

portion of a structure, and beam-to-column web connections. Tests conducted

by Rentschler and Driscoll on beam-to-column web moment connections

subjected to a statically applied bending moment raised concern about the safety

of such connections under a dynamic loading condition [8]. In one of these

tests, failure of a fully welded moment connection was due to an instantaneous

break across the entire tension flange near the welded zone when the applied

load reached 85% of the plastic moment. Clearly this does not indicate good

ductile behavior. Results by Popov [6] on a similar but smaller member

subjected to cydic loading demonstrated a brittle type fracture that was also

unsatisfactory.

Poor results could be traced to a combination of causes such as the

geometry of the connection, location of the groove weld and defects m the

welding procedure or workmanship. These parameters were examined in a series

of reports by Pourbohloul [7]. Variables of these tests were the connection

plate geometry, plate thickness, weld size and the use of a backup stiffener.

Results of these tests improved the knowledge of the failure mode of column

connections so that design recommendations were made to direct the course of

further study.

2

The current investigation was conducted under Fritz Engineering

Laboratory Project 504 funded by National Science Foundation Grant No.

ECE-8320540. The grant was for a program including theoretical studies and

full-scale tests on weak-axis beam-to-column moment connection details [5].

Test setup design began in September 1985, fabrication of the components was

done in February thru May of 1986 and testing was conducted during the

months of June thru November of 1986. Originally, the project was to include

two tests, but due to a premature weld failure in the first test it was possible

to repair the specimen and retest it for a total of three tests.

The intent of the project was to simulate a single-sided beam

subassemblage as would be found on the exterior of a building. See fig. A-1.

The test specimen was restrained in a large fixture, and loading was applied at

the tip of a cantilever arm in cyclic load steps. Each cycle consisted of both

positive and negative vertical deflection increments. The program for the load

steps will be discussed further in the testing section.

Two types of connections were tested. The mam points of Type 1 were:

1.} No backup stiffener

2.) Fully welded around beam flange connection

plate to column web and flanges

3.) Bolted web connection

4.) Beam flange connection plate extended past column

flange tips

The Type 2 connection detail differed from the above m that the beam flange

3

connection plate was not welded to the colurim web, and the beam flange

connection plate was increased in thickness to account for the full-span effect

between the column flanges. See fig. A-2 & A-3 for design details of Type 1 &

2.

The significance of this test is that only a limited number of weak-axis

beam-to-column moment connections have been tested under cyclic loading

conditions. The members tested here are larger than any previous cyclic loading

test of this type. There are some effects of size that can not be scaled up,

they can only be determined experimentally. Once the phenomenon that has

occurred in the full-scale test is understood then a model can be developed to

predict future behavior. Guidelines must be established to insure the safety of

large-scale fabrications so they can withstand severe loading conditions such as

an earthquake.

Attention was focused on the details of the welding procedure, weld

location and size, thickness of the connection plate members, and number and

location of the bolts in the web plate joint. A balance between member size,

ultimate strength, and ductility must be achieved m order to maintain

structural integrity during large distortions into the inelastic range.

4

Chapter 2

Considerations in the Design of the

Specimen

In order to maintain continuity with the earlier work of Rentschler [8] on

full scale connection tests and the flange connection plate detail tests of

Pourbohloul, et. al. [7], a W27 x 94 beam and two columns, a W14 x 174 and

W14 x 257 were selected as typical members in a building frame. Primary

components that must be designed are:

1.) Beam flange connection plate

2.) Beam web connection plate

3.) Welds of connection plate to column

4.) Full penetration weld between the beam

flange and the connection plate.

5.) Location, number and size of bolts m

web joint.

Each of these items will be discussed below:

Item !: The thickness of the beam flange connection plate for the Type 1

& 2 connections was determined by the previous work of Pourbohloul [7]. The

plate was extended three inches past the column flange tips in order to reduce

the stress concentration at the re-entrant corner. The plate was tapered in the

thickness direction at a slope of 1:2.5 in accordance with the Structural Welding

Code [2] fig. 8.10A for butt joints in parts of unequal thickness and offset

alignment. The plate was not . tapered in the width direction. However, the

amount of extension provided was sufficient to permit the path of tension

5

stresses to flow at an angle of 45 degrees from the column flange tips to the

edges of the beam flange as recommended by Pourbohloul [7]. Plate thickness

and the lack of welds between the column web and the connection plate were

the only differences between Type 1 & 2.

Item ~: The beam web connection plate was sized to be 9/16 inch in the

thickness direction, but obtaining plate this size was difficult so the thickness of

the web plate was increased to 5/8 inch to allow delivery of a stock item.

Care was taken to allow a generous radius at the cut out portions of the plate.

Item ~: Weld sizes were originally designed on the basis of the maxtmum

static loads to resist the shear and bending moment components of the applied

vertical load. This was shown to be inadequate in Test 1 because the Poisson's

ratio effect produced through-thickness stresses that caused cracks to develop at

the root of the weld. A detailed description of the through-thickness effed will

be given in the discussion of results. These welds were double fillet all around

the edges of the plate except the column web weld was eliminated in the Type

2 connection, Test 3.

Item 4: The full penetration weld between the beam flange and the

connection plate was per the ANSI/ AWS 01.1-86 code, fig 2.9.1 for single bevel

groove weld butt joint with backing bar [2]. This weld was dye-checked and

ultrasonically tested for the first test. The backing bars were removed and any

irregularites ground down and repaired. This procedure was done to reduce the

problems due to stress concentrations cited in the work of Pourbohloul [7].

Further comments will be made in the discussion of results.

6

Item _!!: Seven one-inch A490 bolts were used in the design to carry the

vertical shear load, but it was not possible to design a single row of bolts to

develop the entire plastic capacity of the beam web. This meant that the beam

flange connection plate had to accept higher normal forces due to the reduced

capacity of the web plate. (Later work by Lu tested the ultimate capacity of

these bolts.)

The specimen was proportioned with the overall length of the cantilever

exceeding the distance in which shear load would affect the final failure mode.

Fabrication was done in Fritz Lab and the proc~ss was closely monitored. Care

was taken to orient the rolling direction of the connection plates with the

longitudinal axis of the beam. Failure to do this was cited as a problem in

earlier testing. Bolt torque was checked by turn-of-the-nut method.

The fabrication quality of the finished specimens was good and probably

better than the average field craftsmanship. The welds were of proper size ·and

did not have porosity, surface cracks or inclusions visible to the naked eye.

Ultrasonic examination of the groove weld revealed some small voids and/or

inclusions but these were not large enough to require repairing the welds.

These inclusions were documented. Comments regarding the fracture mechanics

approach to the weld failure will be made later. A heavy mill scale on the

surface of the connection plate may have affected the depth of penetration and

this will be discussed in the results section.

Additional aspects of the specimen designs were based on results of other

7

investigations. A properly designed and fabricated connection will exhibit

ductile behavior and failure away from the groove weld of the connection plate

to beam flange joint. Popov reached some fundamental conclusions as to the

result of his work. The load deflection hysteresis loops for a steel cantilever

beam and connection are highly reproducible during repetitive load application.

This fact implies that such an assemblage is very reliable and can be counted

upon to absorb a definite amount of energy in each cycle for a prescribed

displacement. The ability to withstand severe repeated and reversed loading

seems to be assured for properly designed and fabricated steel structures; their

intrinsic energy absorption capacity IS large [6]. This energy absorption

capacity, as measured by the s1ze of the hysteresis loops, increases with

increasing tip deflection until the maximum loop outline is reached.

These assumptions form the basis for companson of test results for

different connection details. Some sort of standard must be established as a

guideline for an acceptable connection design. The ultimate energy absorption

capacity of the test detail must be compared to this standard to determine if it

meets the minimum requirements. For this series of tests the standard was

chosen at three cycles where the load reached positive plastic moment and

negative plastic moment based on recomendations by a European committee [9].

Chapter 3

Considerations in the Design ot the Test Set-Up

As stated above, this test was umque m that full stze members were

cyclically loaded to plastic bending failure. Due to the magnitude and

application rate of the forces required to cause failure, careful design of the

reaction frame was mandatory. Unfortunately, space constraints on the testing

floor and availability of testing frame members restricted the reaction frame to

less than optimum design. The frame used for this test was essentially a two-

dimensional structure with the cantilever beam, column and loading mechanism

in the plane of the reaction frame. See fig. A-4. This greatly reduced the

width of the frame allowing it to be erected in a small corner of the lab, but it

did not allow for cross-bracing. A forty-five degree leg braced one corner only,

while lateral struts stablized the frame to the building wall. The entire frame

was bolted together so that it could be taken apart to change the test

specimen.

The loading forces and reactions were modeled using STRUCTR--a

Structural Analysis Program developed by Driscoll and students at Lehigh

University [4]. The program inputs the number of members, joints, type of

restraints, member properties and material constants. Using the output from

the program, the deflections and forces at all ends of the members can be

determined.

Another factor m the design of the test 1s the loading system for the

9

beam. Since the test was to simulate dynamic loading conditions as much as

possible, hydraulic power was chosen as the prime source for a displacement

driven test program. Two hydraulic actuators operating in series (i.e., one in

compression the other in tension) were needed to reach the failure load of the

specimen and not overload the capacity of the actuators at their extended

length. The units used in this test were 8-inch bore by 24-inch stroke

T. J. cylinders manufactured by Aero-Quip. Pin connections at both ends of

each cylinder allowed displacement of the beam and rotation of the cylinder

without binding the piston to the bore. The hydraulic pumping system used

was an Amsler swing arm pumping unit which has exceptional flow control and

can maintain a steady pumping pressure almost indefinitely. Direction of travel

of the actuators was changed by using a Moog servo valve with a remote

control at the Amsler control panel. After some initial start-up problems the

system performed almost flawlessly.

After the connection ·details had been designed, the reaction frame built,

and the hydraulic system put together, the next major phase and probably the

most critical was the acquisition, storage and reduction of the test data.

Instrumentation line-up for Test 1 had 31 rosette strain gages, 8 linear gages, 7

dial indicators, 2 linear variable displacement transformers(L VDTs), a load cell,

and 2 rotation gages. There were 101 channels of strain readings to be

recorded at each load step. Four of the strain gages were wired together m

series and read manually by an operator. These four channels were calibrated

at a specific point along the length of the beam away from the yield zone to a

50 kip load cell. Using this calibration and extrapolating the load range

10

(assuming linear behavior) the load could be read versus a prescribed

displacement. Subtracting these four gages from the total left 97 remammg

channels that had to be recorded. A micro-based data acquisition system was

used for this purpose. Unfortunately, the system proved to be the weak link in

the operation. Unreliability, problems with the software, disk drive errors,

slowness, and limited screen visibility created many headaches. Some file

security was achieved by linking the system to an independent m1cro and

transferring data files to its hard disk once or twice a day. Even with this

backup, some files were lost. Counting all the delays and breakdowns, at least

two weeks were lost during the course of the test.

In addition to the strain gages, the other items of interest had to be

recorded manually. Dial gages, LVDTs, load cell readings and various book­

keeping information such as load number, date and time had to be written

down in specially formatted tables for each load step in the cycle. Some

readings such as load and displacement were checked at the beginning and end

of each recording cycle to make sure the values had not slipped due to yielding.

This was especially critical at displacements in the inelastic range. The entire

recording process took three to five minutes (depending on the size and skill of

the crew) and to complete one load cycle could take four hours.

See fig. A-5 for the location of the dial gages and L VDTs. Dial gages

were used to measure bolt slip at the edges of the web connection plate, the

movement of the web connection plate at the center of the column web and the

Poisson's effect at the column flange tips where the horizontal connection plates

11

were located. The L VDTs were used to measure the displacement of the beam

tip (this was called the "criterion" measurement since it controlled the test) and

the sway at the top and bottom of the column. Column sway was measured to

calculate the angular rotation of the column centerline which contributed to the

beam tip displacement.

Photographs were taken of the overall test setup and at vanous stages of

interest such as maximum load or fracture of a weldment. As mentioned above,

an attempt was made to measure the rotation of the beam with respect to the

column· using some very old rotation gages but these did not work.

Unfortunately, this critical angle of rotation could not be determined reliably.

12

Chapter 4

Results of the Test Program

The programmed loading sequence was applied in increments of load and

displacement over a range of cycles that represented a prescribed percentage of

the maximum yield load. Each cycle was repeated three times in order to

simulate the energy of an earthquake. The peak amplitude of each group of

three cycles was chosen to be one fifth, two fifths, three fifths, four fifths and

equal to the amplitude required to cause plastic moment. The plan was to

work gradually up to the plastic moment and then repeat this cycle three times.

If the specimen were still intact then three more cycles of twice the amplitude

of the plastic moment were to have been applied. However, as will be seen this

procedure was somewhat academic.

Test !--This test was conducted on the Type 1 connection( see fig. A-2

for details) which was welded along the flange plate connection to column web.

The load was applied as described above, see fig. A-6. As shown in the plot

the connection failed in the eleventh cycle at load step 158, which was

approximately 86% of the required plastic load. Failure was by a sudden

fracture of the left fillet weld in the top flange connection plate. The fracture

occured on both the top and bottom fillet welds along the column flange

portion of the plate and around the corner through about half of the fillet

welds along the column web. The initiation point for the crack appeared to be

at the column flange tip and it then worked slowly back along the inside face

of the flange when the weld was m tension. When the weld was in

compression the crack appeared to close. Fracture mechanics states that when

the crack driving force exceeds the crack resistance strength, unstable growth

occurs. As the load was increased to higher levels, the driving force increased

as the crack grew and the stress intensity level was raised. Crack growth

resistance increases with small increments to crack growth so the resistance was

able to keep up with the early stages of crack extension. Ultimate crack

growth resistance IS a finite material property so when the limit is exceeded

there is nothing left to resist fracture; therefore brittle (sudden) failure results.

Failure in this test was at a load about 14 percent less than expected and

the major components had undergone little if any yielding. A decision was

made to repair the welds and resize them considering the through-thickness

effect of Poisson's ratio. This will be discussed in greater detail in Chapter 7.

Thus the fractured 13/16 inch fillet welds were burned out and replaced by 1

inch fillet welds. The specimen was put back in the test frame and prepared

for testing.

Test ~--The second test on the repaired specimen gave much better results.

The connection had adequate strength to develop the full plastic moment of the

cantilever beam. There was also adequate ductility to permit three cycles of

reversed plastic moment. See fig. A-7 for a plot of the load versus increment

number. After completing the program of controlled loading cycles, a final

monotonic load cycle was applied m order to determine the maximum

deformation capacity of the speCimen. Loading in the negative (downward)

direction was halted when a crack initiated in the groove weld connecting the

beam tension (top) flange to the connection plate. Loading was reversed,

14

because the crack area could resist compressiOn as the crack closed.

Displacement in the reversed (positive) direction was applied until cracking

occurred in the fillet welds along the column flange on the lower right hand

side. The maximum load reached was about 122% of the calculated plastic

moment. The maximum displacements were about +168% and -195% of the

observed yield displacement.

Examination of · the column flange fillet weld crack showed that it was

quite similar to the failure in Test 1. Again the crack appeared to start at the

column flange tips, perhaps due to a poor penetration of the root pass in this

area. However, this time the fillet welds displayed quite a bit of ductility

(yield lines can be seen along the weld surface) and held together for large

strain rates.

The crack in the tension flange full penetration weld can be shown to be

initiated by an elliptical flaw in the middle of the weld zone. The presence of

this flaw generates a high stress field around the perimeter of its boundary,

these high stresses produce localized cleavage failure. This can be seen as areas

of flat fracture. As the the crack progresses along the width of the beam

flange its mode of failure becomes slanted indicating a more ductile or shear

failure. The initial flaw may have required several cycles to grow large enough

to induce a critical stress field but once it reached this limit the growth became

unstable.

Test 3--The third test was conducted on the Type 2 specimen (see fig.

15

A-3) in which the fillet weld along the column to the top and bottom

connection plates was eliminated. This had the effect of creating a bridge

between the inside faces of the column flange, making the welds along this area

even more critical. For this test a heavier column section, a W14 x 257 was

used. The planned loading sequence was applied. See fig. A-8 for a plot versus

load number. As can be seen from this plot, the connection detail was able to

achieve the theoretical plastic moment in both the positive and negative

direction. The goal of three cycles at plus or minus plastic moment was not

met, so the Type 2 connection did not meet the full ductility requirement.

Again cracks were seen to initiate in a fillet weld along the column flange,

this time it was the lower left corner weld. The crack was observed to grow

when the connection plate was held in tension. During the start of the third

cycle of the plastic moment loading sequence the crack fractured suddenly. The

results were mixed: the desired load was achieved, but the ductility was not

acceptable.

In reality the ultimate plastic moment was not reached in this test either.

This can be seen by comparing the load versus deflection curves for Test 3 with

Test 2. See fig. A-9 and A-10. In Test 2, the. curve "flattens out" as the

ultimate plastic moment is approached, in Test 3 the curve is still climbing as

can be seen by the relatively steep slope. The theoretical plastic moment was

achieved in Test 3 but this is calculated based on a yield strength of 50.0 ksi.

The beams used in these tests had a yield strenth of 58.0 to 60.0 ksi hence the

ultimate load should be higher than the theoretical by about 20 percent. This

16

goal was reached in Test 2 as the ultimate load was about 22 percent higher

than the value calculated using a yield stress of 50.0 ksi. The ultimate value

was higher than the value calculated using a yield stress of 60.0 ksi even

though yielding had not progressed through the entire beam web. It is doubtful

that a moment connection with a bolted web joint will ever develop the full

plastic moment simply because of the high strain rates required to plastify the

web. Something usually fractures in the top flange connection plate assemblage

before the plastic zone extends completely through the web.

17

Chapter 5

Test Data Management

As discussed previously, each of the three tests generated huge volumes of

data. In the past, the general procedure is to use existing software on the data

acquisition system to calculate the desired stress-strain results. This can be

done on certain systems. The software for these machines is somewhat limited

and their fixed output format makes application of some other post-processing

device difficult, i.e., if one wishes to obtain plots of various stress results.

A different strategy was used for this project. A micro with a hard drive

disk was connected to the data system via an RS 232 port running an

interfacing program. This hook-up allowed data file transmission of the

automatically recorded results to an independent hard drive for later post-

processing. This streamlined the data reduction effort while adding much

needed file security to backup the delicate dedicated disk drive. It was no

longer necessary to run the built-in stress-strain processing software and view a

paper printout. Once the raw data files were on the hard disk they could be

neatly copied to floppy disks for convenient storage and transportation.

The raw files had to be reduced and grouped according to the strain

rosettes or linear gages that corresponded to the actual channels read. Also,

there were 13 or 15 manual readings for each load step of each test. These

manual readings had to be combined with the automatic readings taken at the

same time. A scheme was devised to do this using a micro running a Basic

language program. Three programs were used to reduce, collect and combine all

18

the information for a particular load step. First, program DATIN took a raw

data file and compressed .it into a format that could be combined with the

manual data. This program removed all the headings, titles, page numbers, etc.

that were present in the formatted printout of raw data and converted them to

a list of channel numbers and the corresponding strain at each load step.

Second, a program called MANIN was used to input the manual data at each

load step. The program prompted the user for the given manual reading, i.e.,

Dial Gage No 1 or LVDT No 2 etc. for each load step which was then written

to a file for the manual data. Each file consisted of ten to fifteen load steps;

care was taken to match the starting and ending number with the corresponding

number for the raw data file from the data acquisition system. Test 1 had 159

load steps, Test 2 had 194 load steps, and Test 3 had 239 load steps. One can

easily see that many data files were generated and record keeping was very

important.

The third program was CO MDA T and logically enough this program

combined the compressed electronic data file with the corresponding manual data

file. In this manner, all the raw data records for each load step were

consolidated in a single list. All this sounds very simple but considering all the

little mix-ups and acquisition system operating problems the above operation

took months.

Once all the data files were generated the actual post-processing could

begin. The scheme behind the data processing was this:

1. Input a setup file with test parameters describing location of gages,

19

channel numbers, material properties and desired items to be

calculated.

2. Take all raw data stored and subtract the zero reading to get the

difference.

3. Take the difference readings required for a particular rosette/linear

gage and transfer them to a subroutine to calculate the stress-strain

relationships.

4. Repeat step 3. for manual data

5. Store the load number, date and time at the head of a data file

created for each load step. Then store the differences for strain and

manual readings. Then store the calculated results, the linear stress

results and last store the calculated results for each manual reading.

For each load step there were approximately 400 items (depending on the

number of rosettes per test) to be stored. This was done by opening a direct

access file and writing the list of items to it for each individual load step. The

program that did all this was called DPROC. The program could be run on a

micro with 640 K of central memory.

The key point of the direct access system is that now all the results are

stored at a particular location in a direct access file. Any result can be

20

retrieved by another program designed to select a location in the first load step

stored and then proceeding to the same location in the next load step. This

can be achieved by setting the file pointer to a pre-calculated value depending

on the result desired and simply incrementing this pointer value by the number

of records stored in each load step. The number of records is a preset constant

and this value is added to the initial pointer value for each load step until the

file has been read to the end. The desired result is then stored in a temporary

array for viewing purposes.

In this manner, a stress value at a particular location on the specimen can

be collected for each load step of the test. Once these values have been

collected in an array they can be plotted on the screen of the micro-computer

using a graphics subroutine. Any desired value can be displayed. For added

flexibilty a value could be plotted as an ordinate with the abscissa the load

step number or the value could be plotted on the abscissa with the ordinate

being the load value in kips {eg. load versus displacement). The program that

did this was called DPLOT. Once the graph has been displayed on the screen

the viewer has the option to create a file for an X-Y plot. This file can be

sent to a plotter for hard copy. The plotting software used for this research

project was the AUTO_ GRAPH [3] program on the CAE lab mini-computer at

the Fritz Engineering Laboratory. This program creates high quality plots and

allows the user to plot multiple functions, i.e., two stress results can be

displayed on the same plot. A Hewlett Packard pen plotter was used for all

the plots found in this thesis.

21

The two post processing programs and the data reduction programs were

written and developed by Dr. George C. Driscoll and students at Lehigh

University. The use of these programs on the micro-computer provided

tremendous flexibiliy and power m the data reduction which allowed an in-depth

study of the experimental data.

22

Chapter 6

Computer Modeling

The popular · technique in computer modeling the past several years has

been the finite element method, FEM, using one of the package programs such

as SAPIV, ADINA, ANSYS or some other program in vogue. As is common

knowledge, this method discretizes the structure as a series of small elements.

The number of elements is limited more or less by the size of the computer.

But really, the program is limited by the patience of the programmer. In order

to make an accurate analysis, one must use more and more elements. Hence

the model builder tends to limit the size of his model by making assumptions

or restricting the model to a two-dimensional plane. What one generally

achieves is a very detailed analysis of a small portion of the original structure.

Even with this approach, FEM generates more data than the average engineer

can interpret. A worse effect is that one tends to overlook behavior caused by

the three-dimensional loading of a structure. Important loads, stresses and

reactions are often overlooked by a simplified model.

The intent of this research project was to use a more approximate analysis

technique, matrix analysis, but use a more complete three-dimensional model.

In this manner, the entire structure could be represented and analyzed using the

direct stiffness method program STRUCTR. Some modifications to the program

were necessary for the application to this connection detail. STRUCTR is

generally used for the analysis of large span structures, I.e., bridges, building

frames, truss members, etc. where the length-to-depth ratio is large. This

implies that the major component of the element stiffness matrix is distortion

23

due to bending{flexure). In a beam-to-column connection fabrication, the length­

to-depth ratio of the members is about one or less. Shear distortions account

for a large percentage of the overall element distortions. The program

STRUCTR was modified to include the shear distortions' in the element stiffness

matrix. The revised program was called STRSHR.

The development of the model was actually simple. Moment of inertia

section properties were calculated for bending flexure about the strong and weak

axis of each connection plate member and a portion of the column web.

Effective shear area was calculated for shear distortions and polar moment of

inertia for a thin rectangle was calculated for torsional rigidity. Section

properties for the W27 X 94 beam were found in the AISC handbook [1]. The

column web was modeled as two separate fixed end beams to include the

noticeable deflection of the centerline of the column web at the intersection of

the beam web connection plate. See fig. A-ll for a layout of the model.

Members 1 & 2 and 3 & 4 represent the column web with points 1, 3, 16, and

18 being fixed restraints provided by the heavy column flanges. Members 5 &

6 and 7 & 8 represent the bottom and top connection plates respectively.

Members 9 & 10 represent the horizontal connection plate and members 12 &

13 represent the vertical connection plate. Points 4, 6, 10, 19 and 21 are fixed

supports provided by the column flanges. Member 11 IS the cantilever beam

with the load at point 14. Members 14 thru 25 are dummy rigid members

with very high section properties to transfer reactions from one active member

to another while maintaining geometric relationships between the members.

Joints 7 and 22 are pinned allowing Z axis rotation becauses no relative

24

moment can be transferred through the weld. It is assumed that the weld joint

transfers only axial force and vertical shear. Joints 9 & 15 are also pinned

about the global Z axis. Joint 9 represents the center bolt of the lower group

of three bolts while joint 15 represents the center bolt of the upper group of

three bolts. This was done to provide an average moment caused by the

leverage between the upper group of bolts and the lower group of bolts on the

web connection plate.

The model included twenty five elements and twenty three node points,

but it fully described the three-dimensional nature of the connection providing a

valuable insight into the distribution of the member forces and external

reactions for design purposes. An exploded force diagram illustrates this point

(see fig. A-12). Traditionally, connections are designed assuming that the

vertical web carries all the shear force and the horizontal beam connection

plates carry only normal bending stresses uniaxial to the direction of the beam.

Examination of fig. A-12 shows that less than 50% of the vertical shear is

carried by the web plate and a secondary moment is carried by the top and

bottom connection plates due to the component of the vertical shear that is

transferred into these members. Thus, even a simple model such as this one

provides new information regarding the force distribution in the connection

members. The model does have limitations in that stresses due to the through­

thickness effects of Poisson's ratio are not included. A comparison of the model

to the actual stress distribution will be discussed in the next chapter.

25

Chapter 7

Discussion of Results

This project was unique in that the theoretical analysis was limited, usmg

only a matrix analysis program to solve a simple geometrical model. On the

other hand the the amount of data collected and assimulated for computer

based retrieval was voluminous. The key to making sense of the data recorded

was knowing what to look for and then spot this trend while interpreting the

data. Probably over a hundred stress versus load value or number plots were

made in order to achieve the few simple observations that will be put forth.

Bolt slip IS one of the pnmary causes for redistribution of the forces in

the connection plate members. Bolt slip begins imm~diately upon application of

the first load cycle and parallels the direction and magnitude of the applied

load. See fig. A-13. As the flanges begin to yield due to plastic flow the bolt

slip becomes at least ten times greater than the elastic range value. A

thorough discussion of the consequences of this behavior will be given later.

The top and bottom flange connection plates are subjected to bending about the

global X and Y axes. Bending about the X axis (YZ plane) approximates the

shape of a fixed end beam with a single concentrated load at the center span.

The plate is bent in reverse curvature as the heavy fillet welds on each end

truly act as fixed supports. This behavior was noticed when analyzing the

rosette results in this plane. See fig. A-14. It can be clearly seen that the

stress at the exterior rosettes must be opposite m s1gn to the center rosette

gage because the exterior rosettes are outside the center zone defined by

inflection points a and b. Although this IS a rather simple and straightforward

26

observation, it took some time to deduce. First the normal stress in the global

Z direction had to be interpolated to the neutral axis of the XZ plane bending

so the normal stress caused by Y axis bending would not influence the X axis

bending stress. Then it was observed that the Z direction normal stress

reversed sign between points 1 & 2 and 2 & 3. Finally, the relative

proportions of the stresses were of such a ratio as to suggest fixed end

conditions. This makes sense when considering the top flange as a thin plate

member with a length-to-depth ratio of about eight to one. The heavy 1-1/8

inch double fillet welds act to restrain rotation of the plate ends creating a

truly fixed condition.

A comparison of the stress distribution with those predicted by STRUCTR

1s shown in Table A-1. If one considers only the length of the top connection

plate that is restrained by the fillet weld to be effective in development of X

axis bending stress, then the moment of inertia at the column flange junction is

reduced by a factor of 1/13.80 (length of weld/width of plate). This has the

effect of increasing the bending stress at this point by a factor of 1.40. This

helps bring the predicted stress more in line with the actual stress along the

column flanges. Another factor to be considered is the through-thickness effect

due to Poisson's ratio. This subject has been examined in a series of reports

by Pourbohloul [7J and these stresses could be superimposed on the results of

STRUCTR. Finally, the shear stress distribution assumed by STRUCTR in the

web connection plate underestimates the actual vertical reaction in the center of

the flange plate. This probably accounts for the fact that the measured

bending stress in the flange plates is higher than the predicted value.

27

Bending about the Z ax1s (XY plane) more closely approximates single

curvature beam bending with a distributed load over a central portion. See fig.

A-15. This was deduced by observing that the Z directed normal stress always

maintained a gradient of the same sign for a given loading direction between

rosettes Rl2 & Rl5, Rl3 & RI6, and Rl4 & RI7 which implies a simple beam

deflected shape with pinned ends. See fig. A-29 for location of rosettes. This

can be supported by observing the relative deflection of the tips of the column

flanges. See fig. A-16. This plot shows that the column flange tips move in

and out in phase with the applied load (ie. as the load is applied upward the

beam action pushes the top column flange out and pulls the bottom flange in).

Another reason for supporting the single curvature hypothesis is the fact that

the horizontal connection plates are not welded along their entire length so

there is no restraint over a portion of their cross section.

The input to STRUCTR was modified to account for the single curvature

bending effect of the horizontal flanges by releasing the global Y rotation at the

node points representing the column flange attachment points. This was done

at node points 4, 6, 19 and 21 on the connection model. The revised

connection model was called Model II and had the effect of increasing the stress

closer to levels measured. Members 14 and 17 were also changed to simple

truss members so they would transmit no bending to the column web. This

more closely approximates the actual condition since the top and bottom

connection plates are not welded to the column web.

An interesting comparison can be made between the measured XY shear

28

stress with that predicted by STRUCTR results on the top flange, see fig. A-17.

The shear stress is plotted at R12 & R15 versus length along the flange. The

shear stress is assumed to increase over the constant thickness portion of the

plate and remain constant as this plate decreases in thickness. Integrating this

stress over the length of the plate gives a resultant shear load for this plate

element. Comparing the STRUCTR prediction at Load Step 146 (P = 125.0

kips) which gives a horizontal shear resultant of 172.8 kips versus the resultant

by integration of 182.9 kips yields an difference of 10.1 kips or 5.5%. Finally,

a comparison of the column centerline deflections shows that the values

predicted by STRUCTR at high load steps were within +/- 0.005 inch to those

recorded. The measured value was 0.010 inch versus 0.014 inch predicted.

As stated previously, the traditional assumption regarding the distribution

of shear stress in the beam web does not apply in the vicinity of the bolted

joint. The bolted joint between the beam web and the vertical connection plate

fails to transmit 100% of the vertical shear and bending stress in the web

section. This is due to the bolt slip that occurs immediately on cyclic loading.

See fig. A-18, A-19, and A-20 for a comparison of shear stress between rosettes

R1 & R2, R5 & R6, and R8 & R9. This clearly shows that the upper and

lower pairs of rosettes, R1 & R2 and R5 & R6 respectively, are influenced

dramatically by this slip action while the center pair, R5 & R6, show relatively

equal values. This is expected because there is no bolt slip at the level of the

neutral axis. See fig. A-30 for a location of the rosettes in the web section.

One can see the dramatic drop m normal stress due to the slip action

29

looking at the normal stress distribution. See fig. A-21 and A-22 that show a

comparison of X direction normal stress for rosettes Rl & R2 and R8 & R9.

No appreciable stress is transferred until the strain becomes so large that· the

bolts actually "bottom out" in their holes. It can be seen that once a bolt

becomes locked m one direction it can resist no load m the other direction. If

the strain rate m the other direction were so great that it caused the bolt to

travel the entire clearance tolerance to lock up on the opposite side, only then

would it be able to develop any force in this direction. The unresisted normal

stress and the percentage of web shear not transferred through the bolts must

flow to the connection plate flanges. A review of the angle to the principal

stress for Rl and R8 shows this effect (see fig. A-23).

Another interesting effect is an unequal distribution of bending normal and

shear stress about the neutral axis for higher load levels. Elementary beam

theory states that these stresses should be equal but opposite for bending stress

and equal for shear stress at the same distance above and below the neutral

axis of the member. See Table A-2 for a companson of these stresses between

rosettes Rl and R8 which are both 9 inches from the neutral axis. The

stresses for rosette Rl are significantly higher. See fig. A-24 for a location of

the rosettes on the bottom flange.

The top flange attracts more bending stress which consequently raises the

stress level in the upper half of the beam. A hypothesis which deals with the

geometry of the flange connection plates will be presented to explain this

phenomenon. The top connection plate is about one inch higher than the top

30

beam flange while the lower beam connection plate is flush with the bottom

side of the beam flange. This has the effect of making the connection plate

assemblage stiffer above the beam neutral axis than below (the beam neutral

axis is below the neutral axis of the connection plate). The resultant normal

stress in the beam is thus slightly higher on the upper side than the lower side.

Looking at Table A-2 the bending stress at R1 is greater than R8 for loads 15,

57, 106 and 155. The measured stress brackets the calculated stress values for

loads 15 and 57. At load 106 the top flange is very near yield and the

calculated stress is slightly greater than the measured. See fig. A-25 which

shows a comparison of stress at R20 of the top flange and R29 of the bottom

flange. At load step 155 the top flange has yielded and the stress at Rl is

still greater than R8 but now less than calculated. At load step 214, both

flanges have yielded and the bending stresses are about equal, but much less

than calculated.

This behavior shows how the bending stress redistributes itself in the web

due to yielding and bolt capacity. At high strain rates approaching the plastic

moment of the beam, the bolts have reached their maximum capacity to carry

load and the bending stress in the web becomes equalized between the top and

bottom halves. Once this limit has been reached the additional amount of

stress that would normally be carried in the web by a welded joint must be

transferred to the beam flanges. This hypothesis violates the plane sections

remain plane assumption of simple beam theory and St. Venant's principle does

not apply.

31

.;

In comparing the measured shear stress with the calculated shear stress,

the value measured at Rl is again higher than R8 for load steps 15, 57, 106

and 155. The measured values bracket the calculated value at this location.

At load step 214, both the top and bottom flanges have yielded and the

measured value is greater than the calculated value. The shear stress at R8 is

slightly greater than at Rl. The measured shear values are greater than

required to be in equilibrium with the applied load. This suggests that the

yielding action has increased the shear stress above what would normally be

expected in simple beam theory.

Next the measured shear stress in the beam web was compared to the

value predicted by STRUCTR. A plan was devised to integrate the shear

stress over the depth of the web to obtain a vertical resultant which could be

compaired with the applied load. The shear stress value at a particular rosette

was assumed to be the average value for an assigned area. Since the off-center

gages are plus or minus nine inches from the beam centerline and the beam is

twenty seven inches deep, the effective length for each gage was chosen to be

nine inches. This allowed integration along vertical rows of gages: RI, R5 &

R8; R2, R6 & R9 and R4, R7 & Rll. See fig. A-26 for a general description

of the assumed shear stress distribution along the first row of vertical gages,

Rl, R5 & R8. A simple formula was used to calculate the resultant load:

Peale t*h*(Rl + R5 + R8)

where RI IS the shear stress at rosette I

32

t 0.51 inch for the beam web

t 0.625 inch for the connection plate

h 9.0 inch

See Table A-3 for a comparision of the load at three locations along the

beam and connection plate assemblage. It can be seen that the actual load

value is greater just inside the row of bolts than predicted by STRUCTR but

this switches at the column flange junction where STRUCTR predicts a value

much closer to that actually recorded. This also explains why the measured Z

direction bending stress in the top connection plate is greater that predicted.

The unresisted vertical shear in the web plate becomes a concentrated load at

the center of the top horizontal member. See fig. A-27 for a general plot of

the shear stress in the web. Shear decreases as one proceeds from the beam

web into the connection plate and over. to the column web. The STRUCTR

model could possibly be improved by changing the properties for the members

that model the vertical connection plate.

Some final comments regarding the beam flange and connection plate

junction should be made on the distribution of X direction normal stress in this

area. The normal stress was higher on the tips of the beam flanges than the

value calculated by simple beam theory while the stress at the centerline was

less than this value. This pattern has been observed by past reseachers on

weak-axis moment connections because the stiffer column flanges attract more

33

stress than the flexible column web. Also, the normal stress did suffer some

effects of eccentricity going from the thinner beam flange to the thicker

connection plate. The stress was reduced in magnitude between the two plates

but not by as much as it should have been considering the two areas. An

effective moment arm could be deduced for the eccentricity effect using the

given stress results if it were considered important.

The next important topic to be discussed is welding. Test 1 failed early

due to an undersized fillet weld design that did not account for the through­

thickness Poisson's effect. When this condition was included in the calculation

and the weld repaired, Test 2 gave good results well into the range of ductile

behavior. The weld was resized to account for the through-thickness effect by

using a three-dimensional vector addition that included longitudinal, horizontal

and vertical shear. An overall resultant weld size was calculated that could be

divided by the allowable force per unit length of fillet weld to find the required

size. For the case of the Type 2 connection used in Test 3 the calculated fillet

weld size was almost twice the plate thickness, so an alternative method was

used to find the weld s1ze. The throat dimension of the fillet weld was chosen

to be one half the plate thickness, giving an effective weld size equal to the

plate thickness. Therefore, Test 3 had very large welds but they still failed

early before a full ductile cycle could be completed. The problem appears to be

more of a weld type than a weld size. Fillet welds are poor welds for any type

of cyclic loading because there is always a crack initiation site at the root.

The problem is compounded by the direction of the applied load. Extreme

tensile forces are applied to the longitudinal direction of the welds through the

34

connection plate. As this force increases, the Poisson effect tends to contract

the plate in the thickness direction which wants to stretch the weld. At the

same time, a vertical force in the center of the connection plate wants to rotate

the ends of the plate away from the column flanges. Thus to put it literally,

the welds are being stretched and pried apart at the same time. See fig. A-28.

Yield lines were seen to develop in the weld surface during the test.

This effect is worst at the column flange tips where there is a re-entrant

corner. Fracture mechanics predicts the highest stress concentration at this

point. In each of the three tests, cracks were seen to initiate at this junction

in the load cycles well below the yield load. As each following cycle was

applied these cracks spread around the corner and down the length of the weld.

Eventually crack growth was unstable and brittle fracture occurred.

The Poisson effect caused through-thickness forces between the roots of the

welds, creating a plane strain condition at this joint. This condition has been

observed by researchers working on connection details in previous studies. A

full penetration weld will negate the Poisson effect because it will allow the

weld material to expand and contract with the plate material. A sound full

penetration weld will have no "starter" crack to work through the weld material

as does a fillet weld and it will resist the effects of plate delamination that can

be caused by cyclic loading.

Poor weld preparation may have hastened the destruction but the end

result would still be the same. Each of the three tests failed in a different

35

corner of the beam connection. Metallurgical examination did cite a problem

with some heavy mill scale on the base of one of the broken welds which could

have caused poor penetration. This may have reduced the ultimate strength of

the weld, but since the crack had already grown to such a critical length,

brittle failure was unavoidable. Thus the mill scale can not be blamed for the

failure mode but only for the final violent action.

36

Chapter 8

Conclusions

This research has resulted in a simplified analysis of the connection,

experimental results that corrrespond favorably with analysis, design

recommendations based on the analysis and experimental results, and

recommendations for further study.

It has been shown that a simple three-dimensional model was sufficient to

analyze the connection assemblage using a direct stiffness method matrix

analysis program. The analytical and experimental study was able to provide a

description of the significant force distributions in the connection members.

• Biaxial bending stresses in the top and bottom connection plates were caused by a transfer of a vertical shear force component from the beam web.

• These stresses acted upon the double fillet weld on the ends of the plate causing moment about the longitudinal axis of the weld.

• A vertical force component on the fillet welds was also added by the reaction of the vertical shear not carried by the web connection plate; this is about 0.125 P per double fillet weld.

• Shear distortions are not negligible and must be included m the element stiffness matrix for the plate members.

1

Observations from the experiments resulted in the following conclusions:

• Load capacity of a single row of bolts was shown to be inadequate to transfer the shear and bending stresses from the beam web to the vertical connection plate. This was evident from the start of the test by bolt slippage (applied force being greater than the static friction clamping force) until the very end of the test where the plastic flow in the flanges caused a redistribution of the normal bending stress.

37

• The previously discussed Poisson effect caused internal stress between the roots of the welds that also should be included in the design calculations.

• The redistributed stress pattern intensified yielding of the top and bottom beam flanges.

• Flexing of the column flanges and web was observed in each of these two connection designs demonstrating the local effects of a high stress concentration on these members.

A data management strategy was incorporated into the test procedure to

account for all possible items (automatic or manual) to be recorded, reduced,

combined and retrieved for analysis. The micro-computer programs described in

this text provided data files that could be viewed using plotting software

making a detailed stress analysis feasible.

• The results of the STRUCTR model compare favorably with the overall stress distribution found· by experimental strain gage readings except near the bolted joint.

• It was found that the "web carries all the shear" assumption of simple beam theory does not hold up in a fully welded connection detail, as the horizontal plate members carry significant amounts of the vertical shear.

This approximate analysis was shown to be more informative and realistic about

the actual stress distribution than many sophisticated finite element method

efforts.

Certain contributions to design recommendations may be extracted from

the results of this study:

• For large fabrications involving heavy plate members, the combination of horizontal shear, horizontal bending, vertical shear, and vertical bending requires unusually large fillet welds to join the beam flange connection plates to the column.

• One possible recommendation is to use a full penetration bevel weld on the connection plate to the column flange joint (See fig. A-31 ). However, there must be a range of column and beam web sizes where fillet welds can be used to join the horizontal connection plates to the column flanges.

• The fillet weld on the Type 1 connection along the column web will probably be adequate for the small force observed.

• The use of a backup stiffener would decrease the observed flexing of the flanges and web.

Attempts to formulate design recommendations based of this study and the

results of prior investigations, reveal the need for some further studies.

• Further study should be conducted to find the range of structural sizes where fillet welds will be adequate to join the connection plate members.

• Additional investigations are required to determine the influence of the backup stiffener on the ultimate strength and ductility of the assemblage.

• One final topic for further study concerns a suggestion to reduce the stress concentrat~on at the re-entrant corner where crack initiation was seen to start. A "fitted" plate (see fig. A-31 ) may spread the stress out over a wider area and convert some of the shear stress along the inside of the column flanges to direct pull or tension on the edges of the column flanges. In terms of fracture mechanics, this may eliminate the mathematical singularity point due to the re­entrant corner.

Large scale beam-to-column connections of the types tested in this research

project are often thought of as rigid or fully restrained assemblages. In reality, '

each connection is a fabrication involving individual plate members with very

high section properties about the major axis but significantly reduced properties

about the minor axis. When visualized in this light, connections can be

designed using fundamental classical methods once the three-dimensional nature

of the loading system is fully understoo~9and applied to each component.

References

[1] Manual of Steel Construction,Eighth Edition AISC, Chicago, ILL, 1980.

[2] American Welding Society. Dt.t-86 1986 Structural Welding Code. ANSI/ A WS, Miami, FL, 1986.

[3] Wiedorn, P. G., Seiler, K. W., & Racine, J. Auto_ graph User's Guide CAE Lab, Fritz Lab, Lehigh University, 1986.

[4] Driscoll, G. C. STRUCTR--STRUCTURAL ANALYSIS PROGRAM. Technical Report C. E. 451, Fritz Lab, Lehigh Univ., Sept., 1981.

[5] Driscoll, G. C. & Beedle, L. S. RESEARCH PROPOSAL SUBMITTED TO THE NATIONAL SCIENCE

FOUNDATION--CYCLIC BEHAVIOR OF MOMENT CONNECTIONS.

Technical Report, Fritz Lab, Lehigh Univ., March, 1984.

[6] Popov, E. P. & Pinkney, R. B. CYCLIC YIELD REVERSAL IN STEEL BUILDING CONNECTIONS. Journal of the Structural Division 95(ST3):327:353, Mar., 1969.

[7] Pourbohloul, A., Wang, X., & Driscoll, G. C. TESTS ON SIMULATED BEAM TO COLUMN WEB MOMENT

CONNECTION DETAILS. Technical Report FL 469.7, Fritz Lab, Lehigh Univ., Feb., 1983.

[8] Rentschler, Glenn P., Chen, Wai F., & Driscoll, George C. TESTS OF BEAM TO COLUMN WEB MOMENT CONNECTIONS. Journal of the Structural Division 106(ST5):1005:1022, May, 1980. Fritz Lab, Lehigh Univ.

[9] Plumier, A. Recommended Testing Procedure for Evaluating Earthquake Resistance of

Structural Elements. Technical Report, University of Liege, Oct., 1983. European Convention Tech. Com. 13.

Appendix A

Tables & Figures

41

1.4x Oi/o: 0"'3/ 0"'2 Ld. Gage Measured Calc. Calc. 2

No. Line Stress• Stress• Stress* (-0.626) (-0.366) -------------------------------------------------------------

16 1 -3.362 -2.317 -3.243 -0.602

16 2 6.677 4.416 -0.804

16 3 -6.366 -1.617 -2.264

21 1 3.100 2.216 3.101 -0.399

21 2 -7.770 -4.222 -0.184

21 3 1.430 1.646 2.166

67 1 -4.370 -3.120 -4.368 -0.466

67 2 9.600 6.947 -0.661

67 3 -6.343 -2.178 -3.049

66 1 4.804 3.126 4.376 -0.466

66 2 -10.099 -6.966 -0.416

66 3 4.206 2.181 3.064

106 1 -6.990 -4.685 -6.559 -0.457

106 2 15.310 8.930 -0.534

106 3 -8.175 -3.270 -4.678

114 1 6.399 4.680 6.652 -0.449

114 2 -14.266 -8.920 -0.618

114 3 7.392 3.267 4.673

165 1 -10.360 -6.240 -8.736 -0.486

166 2 21.327 11.893 -0.497

155 3 -10.698 -4.356 -6.098

163 1 7.698 6.240 8.736 -0.418

163 2 -18.422 -11.893 -0.603

163 3 11.103 4.356 6.098 Notez Gage Line 1 connects R12 It R16 Gage Line 2 connects R13 It R16 •-These values ksi Gage Line 3 connects R14 It R17

Table A-1: Com parisian of STRUCTR and Test Results

42'

Ld. No.

15

57

106

155

214

Note1 All

Bending Stress R1 R8 Calc.

-11.21 9.97 10.77

-15.06 12.12 14.51

-20.92 15.49 21.79

-24.45 18.60 29.02

-22.85 23.82 36.30

values ksi

Shear Stress R1 R8 Calc.

-3.68 -2.58 -3.37

-5.19 -3.55 -4.54

-8.54 -5.72 -6.81

-12.72 -8.28 -9.07

-15.38 -16.98 -11.35

Table A-2: Comparision of Stresses for Rl & R8

Load Number 155 Applied Load 125.0 kips

Location

Beam Web (Rl, R5 .t RS)

Bolted Joint. (R2, R8 .t R9)

Near Column Web (R4, R7 A Rll)

'Measured Load

(kips)

119.0

88.2

51.8

STRUCTR Prediction

(kips)

125.0

38.8

60.0

Table A-3: Comparison of STRUCTR with Measured Loads

43

Figure A-1: Typical Connection Detail

44

_J ____ _ ---t---1-~

: ~ ~+

I .o

l ~

Figure A-2: Type 1 Connection Detail

45

TYP.

Figure A-3: Type 2 Connection Detail

46

,...~

Figure A-4: Reaction Frame & Test Setup

47

Section AA

Reaction Frame

DC ff3

DQ ff4

Load Cell Gas••

Stationary column for mea•uring frame &way

A

r DQ ff&

DG ff8

A

Te•t Specimen

LVDT *a

LVDT *3

Figure A-5: Instrumentations for Tests

48

~ ..... OQ

= ... II

> I

~ -t""' (/)

0 Q. ~

~ 0..

""" :;· ........, c.c OQ

Ul 0 ('!) < .0 0 c ('!) ...J ::s n ('!)

0' .., o-3 ('!) CIJ ~ -

TEST 1 LOAD AT BEAM TIP

160.0~---------------------------------------------,

80.0

.0

-80.0

11th Load Cycle Fracture of Specimen

-160.0+-----------~--------~~--------~----------~ .0 40.0 80.0 120.0 160.0

LOAD STEP NUMBER

TEST 2 LOAD AT BEAM TIP

200.0 Fracture of Specimen

~ 120.0 o'Q' £::: ... ID

> I

:-:- ~

40.0 (/)

~ ~ -

~ ~

~ J n.

t"" ~ 0

Cll '-"

"' 9: ::::! 0 0 OQ < en 0 -40.0 ID .0 ...J t: ID

\ \

' ' \ \ •t- '

v ::::! n ID v 0' .., 1-3 -120.0 ·-ID rn ~

t-.:) ' ~

\ -200.0 . . .

.0 40.0 80.0 120.0 160.0 200.0

LOAD STEP NUMBER

TEST 3 LOAD AT BEAM TIP

200.0 Fracture of Specimen

~ ~ A

..... 120.0 OQ

c: ..

'1 II>

> I

OD .. .........

40.0 (/) t:""' a.. 0

~ ~ c;,n -· ........ - 0

OQ 0 en < ~ 0 -40.0 ..0 = ..J ~ 0

~ A ~ ~ A A

\ \

~ ' v v v ·- v v n ~

0' ... 1-:3 -120.0 ~ ·-(/l .... ~ ~ ' -200.0 .

.0 40.0 80.0 120.0 160.0 200.0 240.0

LOAD STEP NUMBER

T2 LOAD VERSUS DEFLECTION

~ 120.0 o'ii. c: ., ID

> I

~ ............ Vl 40.0

t""' a.

0 ~ Pl 1:11 0.. -N I

0 0 ~ c( :::a

-40.0 ~ 0 (") .,.. _J c;· ::s

0' .., >-3 ~ -120.0 en .,.. N

-200.0+---------~--------~----~----~--------~ -5.0 -3.0 -1.0 1.0 3.0

DEFLECTION (INCHES)

T3 LOAD VERSUS DEFLECTION

~ .... (IQ

100.0 ~ tD

> I .... ......... ~ (/)

Q_

t'"" ~ .0 0 -Ill 1:11 Cl..

~ I 0 0 < til !:::1 0 til ~ t"l ..... a· ::s

0' -100.0 .... 1-3 til Ill ..... ~

-200.0-'-----f-------+----'----jl-----+-------4 -2.5 -:-1.5 -.5 .5 1.5 2.5

DEFLECTION (INCHES)

lU

~ ~ ... 0 23 OQ

~ ... 23 ID

~ > I ~ ~ ..

21

J1 ·1& 00 1-3 £ ~ ::::0 p c 0 £ b1 1-3 13 ::::0

"' ;s:: .,. 0

10 11 14 0..

~ tf1. 0' .... y Members: & 0 0 g ::l ::l ('!)

Joints: 1"1 j .... ii1 s·

::s

~ ('!)

3 0'" 8 ('!) .... Cll

I ''

"'%j .... OQ c:: "i ID

> I

'"" N .. 1:':1 >C

"'0 0 0.. ~ 0..

"rj 0 ... n

1:11 ~

1:11 9. fill

CTQ ... fill 3 0' ... () 0 :s :s ~ n .,... c;· :s

s:: ~

3 C"" ~ ... Ul

0.27V

\W14x2S7 Column

O.llV t 0.0018VL t.•tv ..;..._ / ~o.oo•ovL

O.liV~~· o.toV:------Top Plate .-. -.:: -J.lOV O.llV O.J.V

•·••v---.~ -, o.oo•ovL 0.0018VL

"-" O.OIVL

0.48V

o.o~~ 0.17VL

0

0

0

0

O.JIV

*053V

0.07VL

o.tW t o.089V O.llV O.OOIIVL

0.0040VL t /_ --- --t.JIV

O.OOIIVL

)O.I6VL

--w27x94 Beam

-2 TEST 3 BOLT SLIP X10

2.0

" •• .. • • " lo:lj . ' ' ' ...

' I I I ... oq •• ~ .. ' ' s:: I " ... " ' "-' '· CD

> • - ' I (/) .... "' w -1.0 ..

:I: u

tJ:I z 0 ~ .........

"" ~ z 0) "0 0

I» t-.... u t":l w 0.. ~ (JQ -4.0 DIAL GAGE 1 I'D ~ Ill w 0 0 ..... :e I'D 0"

DIAL GAGE 2 -------· '"0 i» .... I'D

-7.0 .0 40.0 80.0 120.0 160.0 200.0 240.0

LOAD STEP NUMBER

Gage Lin• 1

Inflection Point a

Figure A-14:

Gage Line 2

(+)I I I

Web Plate

Gage Lin• 3

--Bending Moment Diagram-

-1.0

Fixed End Beam Supports for Top Flange Plate

57

Wl4 x 267 Column

Flange Connection

Plate~

I I I I I I

-1 I I

lf2'7 X 04 Beaa

Di•tributed Load

Single Curvature Bendina

Figure A-15: Simple Span Bending of Top Flange Plate

58

> I .... ~

•

"""' (/) w J: u z

z 0 i= u w _J lL. w 0

-J X10

TEST 3 DEFL. OF COLUMN FLANGE TIPS

1.0~------------------------------------------------,

-1.0

-3.0

-5.0

-7.0

II l I I fl II e, \ I ll'tl

' ' .. ' ' ',, ' '• ' ''•' _,,I 'tel "~ ., .. ~ '-• I I

•'

DIAL GAGE 3

DIAL GAGE 4 -------·

. ' fl .. I I tl I I 1 I

I ~ 1 I

~ •" '. .. ' . I I I l I I

'' I, tl I I

' ' ' ' ' ' I I

"

• h ..

I I I I I I

' ' ' ' I I I

• I

' ' I '

,I I I I I

I I

' I I I I I I I I I I

' ' ' I

' ' • ' ' I I

'

-9.0+-------+-------+-------~------~------~------~ .0 40.0 80.0 120.0 160.0 200.0 240.0

LOAD STEP NUMBER

••••

o.o

Figure A-17:

Seetioa AA

7.83

W14 X 257 Colu11111

A

13.70

Shear Stress Between. Rosettes 12 & 15

60-

X

~ .... (JQ

= .., Ill

> I I-" OD .. ...........

(/) U) ~ ::r' ('!)

......., I»

(/) .... U) (/) .,....

LaJ ~ ... ('!) Q: - en en 1-t:C (/) ('!) .,....

Q: ~ ('!) L!i ('!)

::s ::0 :t: 0 (/) en ('!) .,... .,.... ('!) en -F." t-.:1

3 X10

TEST 3 ROSETTE NO 1 & 2

20.0

. A

10.0 --

.0 ~ ~ ~ ~ ~ ~

,.,

A p. ~ ~ ' ' ' ' ' I I

~I ' ' ' ' J ' ' ' ' J ' ' '

V{v .l J ' I ' ~ ' ' : ' ' '

~ I ' I ' I ' ' ~ l ' J 'I I • ' ' ' I

' ,, , ,,

-10.0 - ROSETTE 1

ROSETTE 2 ------- u v

-20.0 . . .0 50.0 . 100.0 150.0 200.0

LOAD STEP NUMBER

~

,\ I I

' ' ' ' I

'

250.0

1-zj ... (JQ

£::: '1 ID

> I ..... '!? ,......

(/)

en :X: :::r' -('!)

II> (/) .. en (/) .,.... w

~ ..

0:: ('!) N Ul

CIJ t-t:C (/) ('I) .,....

0:: ~ ('!) ~ ('I)

::I I

:::0 (/) 0 Ul ('I) .,.... .,.... ('!) CIJ

CJ1

~ 0)

3 X10

8.0

4.0

.0

~ ~ ~ J

I

-4.0 -~

ROSETTE 5 -8.0 --

ROSETTE 6

-12.0 • . .0 40.0

TEST 3 ROSETTE NO 5 & 6

I ~ I . I

I I I

' I I I I I I I I I I I I I I I

'~ I I I

' ~ ~ ' • ,• ' • • • • • •

'

-------· _.._ . . . 80.0 120.0 160.0

LOAD STEP NUMBER

!

I I I ' I I I I I I

I I I I ~ I I I I

I ' I • ' • I • • I

\ ~ ~ II ,, ',. ~ I

' f

200.0 240.0

l'2j .... OQ t: ""' ID

> I w c ""' ..

(/) ~ en ......_,

::r In

(/) ~ ... (/)

U)· w .,. ... a::: C) In

"' C/.1 ..__ C/.1

c:l (/)

In a::: .... =e

~ In In ::s I ::0 (/) 0 C/.1 In .,. .,. In C/.1

00

F." co

3 X10

TEST 3 ROSETTE NO 8 & 9

15.0~--------------------------------------------~

10.0

5.0 I I

I I I I

I .0 ' ' ' ' ' ' ' ' I ' ' ' ' I I

' ' ,, -5.0

. -10.0 ROSETTE 8

ROSETTE 9 --------15.0

-20.0 .0 50.0 100.0 150.0 200.0 250.0

LOAD STEP NUMBER

~ .... ~ ... II

> I t-:1

"""' .. .......... Vl

t:O ~ ..........

1'0 :;:!

Vl 9-: ::J Vl

OQ L&J r:n 0:::

a> .... I-~ ., 1'0 Vl Ul Ul

t:O .....J 1'0 <( ....

~ ~ 1'0 0::: 1'0

0 :;:!

~ z 0 Ul 1'0 .... .,... 1'0 Ul -~ N

4 X DIR NORMAL STRESS ROSETIES NO 1 & 2 X10

3.0

2.0 ~

1.0

" ,, ,'• • ,.

• " I' I \ 1 ... ,

.0 . \ \ ,, 1\ \ • ' •

I~ I ' I ' ,I I , j\ I ,,

'I I I' I j

\ I " '. .. v ,, ,, 'l " v ' .. 1

, , ...

-1.0

-2.0 ROSETTE 1

u \ ROSETTE 2 -------·

-3.0 . . .0 40.0 80.0 120.0 160.0 200.0

LOAD STEP NUMBER

...

240.0

x1o 3 X DIR NORMAL STRESS ROSETTES NO 8 & 9

JO.O I'Zj

o'Q" ROSmE 8 s:: ... ID

> 20.0 I

-~ ROSmE 9 ------- ~ w .. w ~ ..

(/')

c:o ~ ......, I!) :::1

(/') 10.0 9: :::1 (/')

OQ w -~.

00 0:: 0') ~ 1-~·

.... I!) (/') Ul Ul

c:o ....J .0 I!) c( ~

~ ~ I!) 0:: I!) :::1 0

.. .- ,, ,, , ,, .. " " ,,

' J.

~ ... , .. ... , .. .. , \ ' .. \ • \ \ \ \

,' ' , ,

::::0 z 0 Ul

-10.0 I!) ~ ~ ~ I!) Ul

00

~ 10

-20.0 -'L . .0 50.0 100.0 150.0 200.0 250.0

LOAD STEP NUMBER

~ 0 Uea•ured

Str••• Di•tr. .... oq

= ... ID 0.

> I w CoO .. en .... .... (1)

~ Theoretical_/' m ~ m 0 4

Siaple Beaa '

~

~

::;- 0 0 3 0 :e (1)

c:r .... 0

~ ;-:::1

OQ

!i o. + CTy

0 Str••• at @G)

y

lo:rj .... OQ c: '1 tD Gas• 0

> Line 3 _IL ~..:. I 0 N lllo. R23 I R28 IL 1"'4 .. ::a:~ I R29 0 VI ~ Gas• .,.._

X .,.._ ~ Line a IL ll: IL 0') 1:""' ~ --0 t'l R22 R25 R28 I» .,.._ c;·

6.6 t:l

0 .... o::l 0 Gaa•

R271L .,.._

....L .,.._ 0 Line 1 _ll: R2' ~ _.:... 3 I torj .L ;- 0.6 t:l

OQ

--1 ~

1.6

Note: All dimeneiona are in inches

4 X DIR NORMAL STRESS ROSETTES NO 20 & 29 X10

16.0 ~ ..... , ...

I I ' ~· I ' ' I

= I ' ' I

"" I I ' I ID I '

I I I I > I I I

I I ' I ~ 8.0 I I I (11 ' I ' .. ~

' I I I (/) I • I I I

tD ~ II I I I ~

........._, ~~ " I I I ::s I l I I I I I

0... (/) I I s· I l I l I I I

(/) I l I l I oq w I I I I I I I . I I en a::: .0 .,...

~ ..., t-

00 ~ en (/) I I I en I I I I I tD

__. I I I I I I ~ 4:( l I I I It .,... . , ., •' ~ ::E ~

,, ~ f ~ a::: • ::s 0

::0 z -8.0 ROSETTE 20 I 0 I en ~ I .,...

I .,... ~ I en , N ROSETTE 29 -------· I 0 I

~ ..,I

N ~ -16.0

.0 40.0 80.0 120.0 160.0 200.0 240.0

LOAD STEP NUMBER

1-,:j p .... oq

= Theoretical

"1 Cl)

> , ,

I to.)

~ t -., .,... 71 t h

(!) oq ..... ~ .,...

en c;·

10 ., 0 -. ;:l;l 0

t v

h 71 ~ ·>. -:-

Cll (!) .,... .,... (!) Cll t 0' .., ~

71 h

• (!)

C"'

00 ::r (!)

_l..J Aasumed Shear Stress

~ .... Be- Web Shear Stresa

o. 13.31 o. 5.<i8

I 7 I

I 7 I

1 7

7 1

s.u. •• 78

1

\ o. 8.<i7 o. 5.<i3

0.

/,-----.,.;;;;(u· 83

Note: All values are KSI

Figure A-27: Shear Stress Distribution m Beam Web

70

\ \

Plate Shriak• ia Thicka•••

--\ --·- M

Rotatioa tZ'7• to PZ'7 welcl apar•

Figure A-28: Through-Thickness Restraint on Welds

71

.,, -

~ oq• 0 =

~.:. ... _IL ID

0 > R14 I RU ll: f"4 I ~

~ IR20 ::tl 0 a ... Ul

Liae a IL ~ IL ~ ~ ~

...:t ~ N RlS Rl6 Rl9 I:""

0 (") a.a ~ ~ o· :;:!

0 :;:!

R18~ _J_ t-3 a ...

-~ ~~ 0 Liae 1 I "0

"!j .L 0.6 iii :;:!

--1 ()q ~

1.6

Note: All climellliona are in inches

Rl

71

y

R5

71 Rl

R3 71

1.0

. f R4 71

.... ~·l­

"1~

IL R2

,_ (t\- ~ R8

X 't~

,.1-l... 't'

,l.l.. '1' ... tr- ~ 'r RO

8.0

1.0

R7

1.0

RlO I£ _L 71 -----...--Rll

~~--------------~" t I

Note: All dimenaioaa are in incha

Figure A-30: Rosette Location on Beam Web

73

1.) Pull Penetration Weld

Figure A-31:

3,) ~end ConnectioD Plate P .. t ColuaA Plana• Tip•

., • ':1 •

Proposed Suggestions for Improvement

i4

,..

..J

Vita

The author was born in Sunbury, PA. on June 27, 1952 the first child of

Blair L. and Marion W. Heaton. He graduated from Selinsgrove High School m

1970 and attended Lehigh University, Bethlehem, PA. where he received a

Bachelor of Science degree in Civil Engineering in June, 197 4.

The author was employed by Bechtel Power Corporation, Gaithersburg,

MD., from June 1974 to October 1977, by Mid-Penn Engineering, Lewisburg,

PA., from November 1977 to January 1979 and by Koppers Co., Williamsport,

PA., from January 1979 to July 1985. He returned to Lehigh University in

August 1985 to pursue a Master of Science Degree m Civil Engineering where

he worked as a research assistant on Project. 504.

75